Research On Key Techniques Of 3D Reconstruction For Space Targets Using Sequence Images

Space target detection ability is key symbol of the nation's comprehensive strength and the development level of technology,and taking advantage of sequence images to get 3D models of space target is one of the important content.The involved key techniques in the whole process of establish 3D model of space targets using sequence images were studied systematically,and a rapid 3D reconstruction experimental system of space target based on sequence images was designed and implemented.The work done and the results obtained in this paper were listed as follows:1.Introduce the research progress of deep space exploration.Making a summarize of the methods of 3D reconstruction based on sequence images.Analyzing the research status and development tendency of three related key techniques.2.On account of the complexity and computation efficiency problems of SIFT,SURF algorithm when applied to space target detection,a SCCH descriptor with 32 dimensions was introduced and an improved matching algorithm combining SCCH descriptor and SURF feature was proposed.The experiment results showed that,computational efficiency of the improved algorithm was better than SURF algorithm with the same accuracy.3.Based on the analysis of the characteristics of observated sequence images of space targets and the requirements of image processing,a sparse bundle adjustment method of space target sequence images was designed.The experiment results showed that,this method could determine the position and posture of relationship between the space targets and sequence images stably and accurately.4.Analysis dense matching method of sequence images of space target based on point spread thoroughly,a dense matching algorithm based on patch expansion was designed,and the surface 3D model reconstruction of space target was reconstructed accurately and completely.5.Based on the above key techniques,a set of space target sequence images 3D reconstruction system software modules was designed and both the simulated and realistic datas were tested for the experiments.